Deep generative optimization of mRNA codon sequences for enhanced mRNA translation and therapeutic efficacy

Area of Science:

• Biotechnology
• Computational Biology
• Genomics

Background:

• Messenger RNA (mRNA) therapeutics offer significant potential but are hindered by inefficient protein expression.
• Optimizing mRNA sequences is crucial for improving therapeutic efficacy and reducing dosage requirements.

Purpose of the Study:

• To introduce RiboDecode, a deep learning framework for designing mRNA sequences that enhance protein translation.
• To demonstrate RiboDecode's ability to improve mRNA therapeutic potency and efficiency across various applications.

Main Methods:

• Developed RiboDecode using large-scale ribosome profiling data and generative sequence exploration.
• Validated RiboDecode's predictive accuracy through in silico analysis across diverse genes and cellular contexts.
• Tested in vitro and in vivo to assess protein expression improvements and therapeutic outcomes.

Main Results:

• RiboDecode significantly enhanced protein expression in vitro, outperforming existing methods.
• In vivo studies showed a tenfold increase in antibody response for influenza mRNA and equivalent neuroprotection at a reduced dose for nerve growth factor mRNA.
• The framework demonstrated robust performance across modified mRNA formats (e.g., m¹Ψ-modified, circular mRNA) and cellular environments.

Conclusions:

• RiboDecode represents a data-driven, context-aware approach to mRNA sequence design, shifting from traditional rule-based methods.
• This framework enables the development of more potent and dose-efficient mRNA therapeutics.
• RiboDecode has broad applicability for advancing mRNA-based treatments in various medical fields.